Microsoft has recently rolled out a series of novel security features designed to assist enterprises in tackling the escalating security threats emanating from AI agents. Notably, the Guardian Agents, part of the Sensing Project, have garnered considerable attention. These agents are tasked with overseeing the functioning of diverse AI agents within corporate environments. Microsoft has highlighted that AI-related threats could stem from agents deployed by the enterprises themselves. These agents may display unpredictable behavior and frequently wield substantial privileges, potentially resulting in dire repercussions if they malfunction. To mitigate this risk, Microsoft advises enterprises to set up an autonomous system for supervising and managing third-party agents, thereby segregating the delivery of intelligent capabilities from control permissions. Although this approach may reflect a degree of caution towards products from AI vendors in which Microsoft has invested or partnered, it adheres to the principle of audit independence and underscores Microsoft's business strategy—namely, promoting its proprietary agent security products to enterprises. Furthermore, Microsoft has enhanced Microsoft Defender with new functionalities, including agent detection, dynamic permission modification, and manual authorization for high-risk operations. At present, perceiving agents as untrustworthy internal threats is progressively becoming a widely accepted notion within the industry. The NVIDIA Open Agent Security Platform, of which Microsoft is a founding member, has garnered backing from numerous tech enterprises. Previously, Microsoft also unveiled laptops powered by NVIDIA chips and equipped with built-in agent security control modules.